We examine whether the predictability of future returns from past returns is due to the market's underreaction to information, in particular to past earnings news. Past return and past earnings surprise each predict large drifts in future returns after controlling for the other. Market risk, size, and book-to-market effects do not explain the drifts. There is little evidence of subsequent reversals in the returns of stocks with high price and earnings momentum. Security analysts ' earnings forecasts also respond sluggishly to past news, especially in the case of stocks with the worst past performance. The results suggest a market that responds only gradually to new information. AN EXTENSIVE BODY OF RECENT finance literature documents that the crosssection of stock returns is predictable based on past returns. For example, DeBondt and Thaler (1985, 1987)report that long-term past losers outperform long-term past winners over the subsequent three to five years. Jegadeesh (1990) and Lehmann (1990) find short-term return reversals. Jegadeesh and

Various theories have been proposed to explain momentum in stock returns. We test the gradual-information-diffusion model of Hong and Stein (1999) and establish three key results. First, once one moves past the very smallest stocks, the profitability of momentum strategies declines sharply with firm size. Second, holding size fixed, momentum strategies work better among stocks with low analyst coverage. Finally, the effect of analyst coverage is greater for stocks that are past losers than for past winners. These findings are consistent with the hypothesis that firm-specific information, especially negative information, diffuses only gradually across the investing public. SEVERAL RECENT PAPERS HAVE DOCUMENTED that, at medium-term horizons ranging from three to 12 months, stock returns exhibit momentum-that is, past winners continue to perform well, and past losers continue to perform poorly. For example, Jegadeesh and Titman (1993), using a U.S. sample of NYSE/ AMEX stocks over the period from 1965 to 1989, find that a strategy that buys past six-month winners (stocks in the top performance decile) and shorts past six-month losers (stocks in the bottom performance decile) earns approximately one percent per month over the subsequent six months. Not only is this an economically interesting magnitude, but the result also appears to be robust: Rouwenhorst (1998) obtains very similar numbers in a

People are overconfident. Overconfidence affects financial markets. How depends on who in the market is overconfident and on how information is distributed. This paper examines markets in which price-taking traders, a strategic-trading insider, and risk-averse marketmakers are overconfident. Overconfidence increases expected trading volume, increases market depth, and decreases the expected utility of overconfident traders. Its effect on volatility and price quality depend on who is overconfident. Overconfident traders can cause markets to underreact to the information of rational traders. Markets also underreact to abstract, statistical, and highly relevant information, and they overreact to salient, anecdotal, and less relevant information. MODELS OF FINANCIAL MARKETS are often extended by incorporating the imperfections that we observe in real markets. For example, models may not consider transactions costs, an important feature of real markets; so Constantinides ~1979!, Leland ~1985!, and others incorporate transactions costs into their models. Just as the observed features of actual markets are incorporated into models, so too are the observed traits of economic agents. In 1738 Daniel Bernoulli noted that people behave as if they are risk averse. Prior to Bernoulli most scholars considered it normative behavior to value a gamble at its expected value. Today, economic models usually assume agents are risk averse, though, for tractability, they are also modeled as risk neutral. In reality, people are not always risk averse or even risk neutral; millions of people engage in regular risk-seeking activity, such as buying lottery tickets. Kahne-

We review extensive evidence about how psychological biases affect investor behavior and prices. Systematic mispricing probably causes substantial resource misallocation. We argue that limited attention and overconfidence cause investor credulity about the strategic incentives of informed market participants. However, individuals as political participants remain subject to the biases and self-interest they exhibit in private settings. Indeed, correcting contemporaneous market pricing errors is probably not government’s relative advantage. Government and private planners should establish rules ex ante to improve choices and efficiency, including disclosure, reporting, advertising, and default-option-setting regulations. Especially

I use a vector autoregressive model (VAR) to decompose an individual firm’s stock return into two components: changes in cash-flow expectations (i.e., cash-flow news) and changes in discount rates (i.e., expected-return news). The VAR yields three main results. First, firm-level stock returns are mainly driven by cash-flow news. For a typical stock, the variance of cash-flow news is more than twice that of expected-return news. Second, shocks to expected returns and cash flows are positively correlated for a typical small stock. Third, expected-return-news series are highly correlated across firms, while cash-flow news can largely be diversified away in aggregate portfolios.

We build a model that helps explain why increases in liquidity⎯such as lower bid-ask spreads, a lower price impact of trade, or higher turnover⎯predict lower subsequent returns in both firm-level and aggregate data. The model features a class of irrational investors, who underreact to the information contained in order flow, thereby boosting liquidity. In the presence of short-sales constraints, high liquidity is a symptom of the fact that the market is dominated by these irrational investors, and hence is overvalued. This theory can also explain how managers might successfully time the market for seasoned equity offerings, by simply following a rule of thumb that involves issuing when the SEO market is particularly liquid. Empirically, we find that: i) aggregate measures of equity issuance and share turnover are highly correlated; yet ii) in a multiple regression, both have incremental predictive power for future equal-weighted market returns.

The paper discusses the current state of the behavioral finance literature. I argue that more direct evidence on investors ’ actions and expectations would make existing theories more convincing to outsiders and would help sort among behavioral theories for a given asset pricing phenomenon. Furthermore, evidence on the dependence of a given bias on investor wealth/sophistication would be useful for determining if the bias could be due to (fixed) information or transactions costs or is likely to require a behavioral explanation, and for determining which biases are likely to be most important for asset prices. I analyze a novel data set on investor expectations and actions obtained from UBS PaineWebber/Gallup. The data suggest that, even for high wealth investors, expected returns were high at the peak of the market, many investors thought the market was overvalued but would not correct quickly, and investors ’ beliefs depend strongly on their own investment experience. I then review evidence on the dependence of a series of “irrational ” investor behaviors on investor wealth and conclude that many such behaviors diminish substantially with wealth. As an example of the cost needed to explain a particular type of “irrational”

The paper has also benefited from the comments of the participants at the Chicago Quantitative Alliance spring meeting, Federal Reserve Bank of New York finance workshop, Harvard University Department of Economics finance seminar, MIT Sloan School of Management finance brown-bag lunch and finance seminar, NBER Behavioral Finance working group meeting, and Stanford Business School finance seminar. Errors and omissions remain our responsibility. The views expressed herein are those of the authors and not necessarily those of the National Bureau of Economic Research.

We examine whether a simple quantitative measure of language can be used to predict individual firms ’ accounting earnings and stock returns. Our three main findings are: (1) the fraction of negative words in firm-specific news stories forecasts low firm earnings; (2) firms ’ stock prices briefly underreact to the information embedded in negative words; and (3) the earnings and return predictability from negative words is largest for the stories that focus on fundamentals. Together these findings suggest that linguistic media content captures otherwise hard-to-quantify aspects of firms ’ fundamentals, which investors quickly incorporate in stock prices.

Abstract not found